blob: 1c89d685729bd45e952f4114a0cecbdda6449de9 [file] [log] [blame]
Reid Spencer5f016e22007-07-11 17:01:13 +00001//===---------------------------------------------------------------------===//
2// Random Notes
3//===---------------------------------------------------------------------===//
4
5C90/C99/C++ Comparisons:
6http://david.tribble.com/text/cdiffs.htm
7
8//===---------------------------------------------------------------------===//
Reid Spencer5f016e22007-07-11 17:01:13 +00009
10To time GCC preprocessing speed without output, use:
11 "time gcc -MM file"
12This is similar to -Eonly.
13
Chris Lattnereab819b2009-01-16 19:33:59 +000014//===---------------------------------------------------------------------===//
15
Duncan Sands8f005a62010-07-07 07:49:17 +000016Creating and using a PTH file for performance measurement (use a release build).
Chris Lattnereab819b2009-01-16 19:33:59 +000017
John McCall2ff24892009-09-24 22:03:45 +000018$ clang -ccc-pch-is-pth -x objective-c-header INPUTS/Cocoa_h.m -o /tmp/tokencache
Daniel Dunbar864db012009-12-12 00:56:47 +000019$ clang -cc1 -token-cache /tmp/tokencache INPUTS/Cocoa_h.m
Reid Spencer5f016e22007-07-11 17:01:13 +000020
21//===---------------------------------------------------------------------===//
22
23 C++ Template Instantiation benchmark:
24 http://users.rcn.com/abrahams/instantiation_speed/index.html
25
26//===---------------------------------------------------------------------===//
27
28TODO: File Manager Speedup:
29
30 We currently do a lot of stat'ing for files that don't exist, particularly
31 when lots of -I paths exist (e.g. see the <iostream> example, check for
32 failures in stat in FileManager::getFile). It would be far better to make
33 the following changes:
34 1. FileEntry contains a sys::Path instead of a std::string for Name.
35 2. sys::Path contains timestamp and size, lazily computed. Eliminate from
36 FileEntry.
37 3. File UIDs are created on request, not when files are opened.
38 These changes make it possible to efficiently have FileEntry objects for
39 files that exist on the file system, but have not been used yet.
40
41 Once this is done:
42 1. DirectoryEntry gets a boolean value "has read entries". When false, not
43 all entries in the directory are in the file mgr, when true, they are.
44 2. Instead of stat'ing the file in FileManager::getFile, check to see if
45 the dir has been read. If so, fail immediately, if not, read the dir,
46 then retry.
Andy Gibbsd77a0512012-10-18 15:24:46 +000047 3. Reading the dir uses the getdirentries syscall, creating a FileEntry
Reid Spencer5f016e22007-07-11 17:01:13 +000048 for all files found.
49
50//===---------------------------------------------------------------------===//
Ted Kremenekf4c45b02007-12-03 22:26:16 +000051// Specifying targets: -triple and -arch
Ken Dycka85ce8a2009-11-13 18:50:18 +000052//===---------------------------------------------------------------------===//
Ted Kremenekf4c45b02007-12-03 22:26:16 +000053
Chris Lattner177b1c72008-03-09 01:36:43 +000054The clang supports "-triple" and "-arch" options. At most one -triple and one
55-arch option may be specified. Both are optional.
Ted Kremenekf4c45b02007-12-03 22:26:16 +000056
57The "selection of target" behavior is defined as follows:
58
Chris Lattner177b1c72008-03-09 01:36:43 +000059(1) If the user does not specify -triple, we default to the host triple.
60(2) If the user specifies a -arch, that overrides the arch in the host or
61 specified triple.
Anders Carlsson3de54ff2008-03-13 03:45:48 +000062
63//===---------------------------------------------------------------------===//
64
65
66verifyInputConstraint and verifyOutputConstraint should not return bool.
67
68Instead we should return something like:
69
70enum VerifyConstraintResult {
71 Valid,
72
73 // Output only
74 OutputOperandConstraintLacksEqualsCharacter,
75 MatchingConstraintNotValidInOutputOperand,
76
77 // Input only
78 InputOperandConstraintContainsEqualsCharacter,
79 MatchingConstraintReferencesInvalidOperandNumber,
80
81 // Both
82 PercentConstraintUsedWithLastOperand
83};
84
85//===---------------------------------------------------------------------===//
John McCalld46f7632011-07-28 07:41:22 +000086
87Blocks should not capture variables that are only used in dead code.
88
89The rule that we came up with is that blocks are required to capture
90variables if they're referenced in evaluated code, even if that code
91doesn't actually rely on the value of the captured variable.
92
93For example, this requires a capture:
94 (void) var;
95But this does not:
96 if (false) puts(var);
97
98Summary of <rdar://problem/9851835>: if we implement this, we should
99warn about non-POD variables that are referenced but not captured, but
100only if the non-reachability is not due to macro or template
101metaprogramming.
102
103//===---------------------------------------------------------------------===//
John McCall56ea3772012-03-30 04:25:03 +0000104
105We can still apply a modified version of the constructor/destructor
106delegation optimization in cases of virtual inheritance where:
107 - there is no function-try-block,
108 - the constructor signature is not variadic, and
109 - the parameter variables can safely be copied and repassed
110 to the base constructor because either
111 - they have not had their addresses taken by the vbase initializers or
112 - they were passed indirectly.
113
114//===---------------------------------------------------------------------===//